The U.S. Census Bureau hires hundreds of thousands of temporary workers to conduct the decennial census. It is tempting to speculate how the large hiring event about to occur in 2020 will affect the unemployment rate, especially in the current COVID-19 environment. Looking back at the impacts during the 2010 Census can provide some insights. Caution is necessary because the economic environments in 2010 and 2020 differ in many substantial ways that we do not fully understand. 

The magnitude of the temporary decennial workforce influences the national employment situation when efforts to follow up with every household that has not yet responded to the census questionnaire ramp up, and when they recede. The Bureau of Labor Statistics (BLS) publishes two headline numbers in the monthly employment situation news release: (1) the number of payroll jobs added, and (2) the unemployment rate.

Each month, the Census Bureau shares with BLS the number of temporary and intermittent employees paid during the week of the 12th of the month. The month-to-month difference in these statistics indicates how decennial census employment influences the number of payroll jobs added. For example, the employment situation news release for May 2010 stated, “Total nonfarm payroll employment grew by 431,000 in May, reflecting the hiring of 411,000 temporary employees to work on Census 2010.”  Figure 1 shows the number of net hires in 2010 for the decennial census. There was a large spike in hiring in May, followed by a gradual reduction in employment beginning in June.

However, the impact of decennial census employment on the headline unemployment rate is less clear. This depends on temporary workers’ employment status in the months surrounding their decennial census employment. For example, if all temporary hires were employed prior to working for the Census Bureau, then there would be no effect on the unemployment rate; the number of employed and unemployed people in the economy would be unchanged. The more temporary hires that were unemployed prior to working for the Census Bureau, the larger the decrease in the unemployment rate due to decennial census employment.

We provide the first quantitative analysis of how decennial census employment affects the employment status of hired individuals, and we draw inferences about the impact on the national unemployment rate. We link the 2010 Decennial Applicant Personnel and Payroll System (DAPPS) data to the 2010 American Community Survey (ACS). The large sample size of the ACS (3.5 million sampled households each year) enables the calculation of valid statistics for relatively small subgroups; decennial census employees are one such subgroup. The ACS informs whether a decennial census worker was employed, unemployed or not in the labor force before working for the Census Bureau. We apply these shares to the set of all DAPPS hires, deriving total flows into decennial census employment from employment, unemployment and outside the labor force (Figure 2a). Similarly, we apply the shares of labor force status after separation from the Census Bureau among DAPPS-ACS linked individuals to the set of all DAPPS separations, deriving total flows out of decennial census employment to employment, unemployment and outside the labor force (Figure 2b). Further details about the empirical methodology and the calculations underlying our findings are in CES Working Paper #20-19, found at <https://www2.census.gov/ces/wp/2020/CES-WP-20-19.pdf> [PDF <1.0 MB].

These flows imply that:

• The large hiring surge in May 2010 came mostly from people already employed (40%) or from people who were unemployed (33%).
  
• Roughly 136,000 of the 411,000 people hired for the 2010 Census in May 2010 were unemployed the month before beginning their temporary decennial census employment, which (relative to the counterfactual) lowered the May 2010 unemployment rate by one-tenth of a percentage point.
• Relative to the counterfactual, decennial census employment influenced the unemployment rate by less than one-tenth of a percentage point in every other month during 2010, due in part to the relatively smooth net separations pattern between June and September.

In future research, we plan to replicate this study with 2020 DAPPS and 2020 ACS once the data become available.